National Efforts to Standardize Brain Scan Data for More Accurate Alzheimer’s Risk Predictions

Transcript

Intro: I’m Dr. Nathaniel Chin, and you’re listening to Dementia Matters, a podcast about Alzheimer's disease. Dementia Matters is a production of the Wisconsin Alzheimer's Disease Research Center. Our goal is to educate listeners on the latest news in Alzheimer's disease research and caregiver strategies. Thanks for joining us.

Dr. Nathaniel Chin: Welcome back to Dementia Matters. We're here at the annual Spring ADRC meeting. I've met with some key figures at the National Alzheimer's Coordinating Center, or NACC, the organizing body of this two-day event. I encourage listeners to tune into those episodes, but for now, I would like to highlight one of the key speakers of this meeting. I'm here with Dr. Beth Mormino, an assistant professor of neurology at Stanford University. She uses advanced brain imaging techniques, in this case positron emission tomography, or PET scans, to study brain changes associated with aging and Alzheimer's disease. Early in her career, she worked on some of the initial studies that used PET scans to determine Alzheimer's disease-related changes begin in the brain decades before a person starts to experience symptoms of dementia. Today she's combining brain imaging and genetics to predict brain health trajectories over time. I highly recommend listeners tune into her first podcast in January 2019 on the science of Alzheimer's disease risk. Dr. Mormino, I'm pleased to welcome you back to Dementia Matters.

Dr. Beth Mormino: Thanks for having me.

Chin: Beth, you presented earlier on a project called “legacy”. Can you explain what “legacy” is and why it's important in the context of AD research?

Mormino: Yeah. The “legacy” project, it's actually quite arbitrary. So in 2021, NIA funded some infrastructure called SCAN, and the point of SCAN was to work with sites to implement standardized image collection both for PET and MRI data across the ADRC program. So SCAN’s job was to implement the rollout of this and also to actually analyze that standardized imaging. Before 2021, before SCAN was rolled out, there actually was a fair amount of imaging that was taking place across the ADRCs, but just not in a standardized way so we're choosing to call that the “legacy” data set. So it just means everything that happened before 2021.

Chin: One of the things you mentioned today is that PET imaging has not been commonly shared among the 30 ADRCs. They were doing these scans but just not sharing them and we've heard in prior talks that sharing data allows for larger samples and possibly more confident findings, or even better detection of important but subtler findings. Why hasn't PET data been shared so widely before, and what have the barriers been to doing so?

Mormino: Yeah, so I think that's actually a very complicated question. One, so actually sites have been sharing their imaging data. There's just been no kind of coordinated efforts and no kind of push to have it in place A versus place B. Actually many sites, they go through these fairly extensive contracts where they're working with other institutions to give their data to those folks and that could be, you know, actually a burden in itself for the site. So sites actually do share their data with individual collaborators. Also there's a number of data-sharing platforms that are out there that folks have used. The thing that has been missing though is, you know, we have all these individual sites that have been collecting this sort of data and there's just been no push to put it into one place so it could be unified across the center and importantly linked to the other affiliated data types that are available on NACC, such as the clinical data. There's also links to the genetic data, and eventually probably plasma data as well. So I think that's one part of the answer. The other part is that imaging in the ADRC program, in many ways it's kind of new. So the ADRC programs – some ADRCs have been around for decades. Some are new and imaging has really rapidly evolved over the last fifteen years, especially with PET imaging. In many ways it has taken place but not as a core component of the ADRCs. It's been through other investigators performing imaging that are collaborating with the ADRC and using ADRC participants for separate research programs. That's another part of the answer is that, it's not – it hasn't really been formally incorporated as a core visit within the ADRC program. So I think for both of those reasons, it hasn't really been shared in a unified way and I think that's what we're hoping to change given the opportunity to do what you said, to actually be able to to benefit from this incredibly large, potential data set that's out there, but it just hasn't been pulled together.

Chin: And in acquiring PETs, or when our participants go through a PET scan, they're given a compound, which is also called a ligand, and this binds to the amyloid protein in the brain but it comes in many different formulations. It's not just one thing that's shared across the country. There's many different types and so the ADRCs can be using different ligands than other centers or other studies. Can PET ligands be compared and then why is it important to harmonize that type of data?

Mormino: Yeah, so that's a good point. So for amyloid PET, there is about four or so compounds that are commonly used. Amyloid PET, you know, this technology has been around for a while now, or over ten years for most of the compounds. A fair amount of work has already been put into place to do head-to-head comparisons across these ligands, so the same person will be scanned with two or more ligands and we can confirm that they're very highly correlated. We do think we are getting the same answer across these ligands. There's some details with that, like the rate the range is different. We do need to do some work to put them on a common scale. So one that is commonly used for amyloid is called centiloids. This is a common scale that brings together these different ligands, so that is work that's been done on amyloid PET. With tau PET, it's more complicated. One, it's a newer technology. There are a handful of ligands that are being used, but that work is actually kind of ongoing on how best to combine that data and how to deal with some ligand specific issues. For instance, there's different patterns of what we call off-target binding in these ligands that are different across them. So we're working on ways to deal with that but the ultimate goal would be to do something like the centilloids where it's a common scale that if you have ligand one and there's a number associated with that, you're able to translate it over to what ligand two would have given you. I would say that's part of the developing science for this new technology and I think projects like what's happening in the ADRC will actually help inform these sorts of harmonization efforts with tau PET.

Chin: And once the data is harmonized, then it wouldn't matter so much which ligand you use because you could convert it to the scale and then you would get a fairly standardized answer so that that result would be roughly similar to someone with the same level of disease in another place using a different scanner.

Mormino: Yeah, that's the goal. So that's to be determined, you know, with the tau PET on how best to do that, but yeah, that is the goal. The goal is to have the common scale, so it means the same.

Chin: And I love hearing that there are dedicated participants that are willing to go through multiple ligands – in essence the same amyloid PET scan but using a different tracer. Thank goodness to our participants who are willing to do this so that we can get to this harmonized, standardized data.

Mormino: Oh yeah, absolutely. We would not be here without the participants driving this science forward.

Chin: So for our listeners, especially our participants who do these scans, it isn't as simple as the scan is done and then the data appears and then you can analyze it, Dr. Mormino. I mean there are procedural and logistical factors that really can impact the image data – or the image and quality of the image and the data. Can you highlight some of these factors that you talked about in your presentation here at the conference and why is it important to consider all of these factors when doing an analysis?

Mormino: Yeah, absolutely so I'll bring up one example I touched on during the talk. What we're noticing with the “legacy” data set is there is a fair amount of variability between when the – for a PET scan the participant will get an injection into their arm and then we capture the data in the scanner after a certain amount of time has passed from the injection. What we're noticing with the “legacy” data set is that that time window between injection and the actual scan itself varied a lot across sites and even within sites. This isn't that important for an overall qualitative read of the data but it is important for the quantification that we want to do, in that the longer the delay is actually the higher the value will become so it could be very misleading if somebody was scanned 30 minutes after injection as opposed to an hour after injection. The person with the hour delay will appear as though there's more amyloid in the brain when, in fact, this is just something related to the dynamic of the actual ligand. This is something that influences the data. It actually changes the magnitude, but it's something we can correct for if we know about it. So it's interesting we noticed this in the “legacy” data set and this is actually the exact type of protocol detail that is fixed by standardization. This is really kind of an example of why standardization is important as opposed to what we're doing with “legacy”, which is kind of this post-hoc aggregation of data that was collected off of, you know, different protocols. These are things that we can deal with methodologically but they probably are introducing some noise that we hope to avoid with standardized data sets going forward beyond “legacy”.

Chin: And that would apply to protocols then too, the idea of ideally standardizing how the scan is done for across centers.

Mormino: Absolutely, and that's what the SCAN initiative that I mentioned at the beginning of the podcast, that's exactly what they're doing. They're telling sites, you know, “for this ligand, this is when you start the scan,” and that will be consistent going forward. That kind of level with standardization or protocols just weren't present in our program before 2021.

Chin: Yeah, the details matter, right, so that's a good example of it. I'm also wondering, how do you determine what an abnormal result is versus a normal one, particularly when you're studying people who are cognitively healthy?

Mormino: Yeah, so this is a big question in the field. I think there's two levels to this question. One is how – so we typically for amyloid PET, for instance, we do typically think of this data as you're either elevated with amyloid or you're not elevated. It's kind of a binary system, although there are some interesting nuances with the continuum. That's one level of the results. When the scan is collected, you can have a trained reader look at that scan and say, ‘Oh that scan is consistent with an elevated profile versus not.’ That's one layer. Another layer is, what does that mean? I think the second part of your question is, what does this mean with a clinically-normal, older individual? And that, you know, we don't know the answer to that. The data suggests that it's associated with an increased risk of future impairment down the line, but to my knowledge we have not developed good individual-level prediction algorithms to actually say exactly what that means if you do get that research result of being elevated. This is, I think, one of the most important open questions in our field is, if you're a normal individual and you have a biomarker positive result from one of these research procedures, what does that actually mean for you as an individual?

Chin: And so lastly, I would like you to share about your data because your presentation was fascinating as far as the findings so far in “legacy”. Do you mind highlighting some of the key things you spoke about?

Mormino: Yeah so, to me, the most exciting piece about the “legacy” project – and actually the ADRC program, in general – is it's really the first dataset – well, it's not really a data, it’s almost like a blend, a multi-site blend of various cohorts – that really has a tremendous amount of heterogeneity within it. I showed a slide briefly during my talk showing the clinical dimensions that have been explored. It ranges from MCI due to AD, to MCI thought to be due to Lewy bodies, MCI thought to be due to vascular disease. This is actually quite different than most of the studies in our space. Most studies very narrowly focus on a very specific set of enrollment criteria, which doesn't allow us to understand how these key biomarkers, such as amyloid and tau levels, how they actually behave in the context of clinical heterogeneity. And interestingly, what we're finding in the “legacy” data set is that we do see a fair amount of amyloid positivity across all these diagnostic categories. It's not just the groups that are clinically thought to be Alzheimer's disease. I think it opens the door for further analyses and questions about the role of this kind of coexisting pathologies or the role of these pathologies that are clinically appearing like something else but might actually be Alzheimer's disease. It really – it feels like one of the first opportunities to kind of get at this question of these classic AD markers in the context of clinical heterogeneity.

Chin: Well with that, I'd like to thank you, Dr. Mormino, for being back again on Dementia Matters, and thank you so much for presenting at this year's spring conference.

Mormino: Yeah, you're welcome and thanks for the invitation.

Outro: Thank you for listening to Dementia Matters. Follow us on Apple Podcasts, Spotify, Google Podcasts, or wherever you listen or tell your smart speaker to play the Dementia Matters podcast. Please rate us on your favorite podcast app -- it helps other people find our show and lets us know how we are doing. Dementia Matters is brought to you by the Wisconsin Alzheimer's Disease Research Center at the University of Wisconsin--Madison. It receives funding from private, university, state, and national sources, including a grant from the National Institutes of Health for Alzheimer's Disease Centers. This episode of Dementia Matters was produced by Amy Lambright Murphy and edited by Caoilfhinn Rauwerdink. Our musical jingle is "Cases to Rest" by Blue Dot Sessions. To learn more about the Wisconsin Alzheimer's Disease Research Center and Dementia Matters, check out our website at adrc.wisc.edu, and follow us on Facebook and Twitter. If you have any questions or comments, email us at dementiamatters@medicine.wisc.edu. Thanks for listening.